[prev in list] [next in list] [prev in thread] [next in thread] 

List:       markus-dev
Subject:    Re: A couple of bugs
From:       Severin Gehwolf <Severin.Gehwolf () utoronto ! ca>
Date:       2010-09-24 3:01:05
Message-ID: 1285297265.2548.117.camel () capricorn ! jerboaa ! org
[Download RAW message or body]

Hi,

> > I notice the link seems to be missing from the sandbox instance.
> 
> I've been going over the code, and the link is only missing for
> assignments where Allow Web Submits is True.  Can't recall when/why
> this was decided, but it looks like we don't allow group cloning to or
> from assignments where web submits were enabled.  Does anyone recall
> the reason?

Taking from
http://review.markusproject.org/r/327/ and
http://review.markusproject.org/r/537/ and
http://review.markusproject.org/r/538/

Some of the interesting stuff:

Severin Gehwolf 7 months, 3 weeks ago (February 1st, 2010, 11:04 p.m.)
I'm adding Karen's message for reference: 

--8<--------- 
I may be coming into this discussion late in the game,
but repos that are sometimes used for web submit and sometimes used for
svn commits sounds like a problem waiting to happen.  In particular, if
a group has svn submit for a1 and web submit for a2, they will still
have svn permissions for the a2 directory, and the different semantics
for web submit vs svn commit could get them into trouble. I'm not sure
how this affects what we set up for assignments.  Can we not allow
carrying groups forward if the submission mechanism differs?
Karen
--8<---------

Karen, the key here is to use different group names for different
assignments. If MarkUs doesn't find a group in the database, it'll
create a new record. As a result a new repository would be created.
Mike, is that right?

Mike Conley 7 months, 3 weeks ago (February 3rd, 2010, 9:44 a.m.)

Hm.
This works if Instructors are not cloning Groups forward.  But what if
they are?
                
Perhaps on cloning groups forward, we should take a look at the
Assignment properties to see if external commits are allowed.  If they
aren't, pull the permissions.
                
There's an edge case, where A2 is created before A1 is completed, and
A2 clones A1's Groups, and A2 wipes out the repository permissions...
                
Perhaps we should distinguish between cloning a *Group* forward, and
cloning a *team of Students* forward.  The difference being that in the
latter case, new Groups/Groupings are created, but StudentMemberships
are copied.  
                
Any more thoughts on this?

---- end conversation.

I think this led up to not showing the clone groups forward link when
web submits are allowed. I might be wrong.

> > I have an assignment set up to be no web submit, and work alone I asked students to log
> > in to find out their repo path. For most students, the repo name is their user id, but for a
> > small number it is group_XXXX? This is puzzling, and a bit problematic.
> 
> That's very puzzling.  I don't have an explanation for that one, sorry.  :(

I'm not sure if this review is related, but there have been failing
tests, which have been fixed.
http://review.markusproject.org/r/686/

Don't know if that's relevant. Maybe we'd need to go back to:
http://review.markusproject.org/r/327/ and
http://review.markusproject.org/r/537/ and
http://review.markusproject.org/r/538/

> > There also still seem to be something wrong with translations for
>  error messages.  Has the fix just not made it into the production
>  instance?

Mike, I think Karen meant something different. Her embedded screenshot
showed some wrong/missing translations errors happening. Showing up in
Rails' flash. MarkusLogger failures would only show up in the log
itself? Am I confusing something here?

So, I'm not sure if this is MarkusLogger related... There might be some
other translations missing?!

Re: missing translations. How accurate is $ rake i18n:missing_keys ?
Should that help for those kinds of things?

Cheers,
Severin

[prev in list] [next in list] [prev in thread] [next in thread]